This systematic review and meta-analysis was conducted according to the PRISMA (preferred reporting items for systematic reviews and meta-analyses) guidelines [20]. Two investigators independently searched the PubMed, Scopus, Embase, and Cochrane databases for all eligible studies. Forward citation tracking was also performed to identify additional relevant studies. The keywords used in the search were “invasive candidiasis,” “candida,” “fluconazole prophylaxis,” “preterm infants,” “very low birth weight infants,” “VLBW,” “extremely low birth weight,” and “ELBW.” Two researchers independently searched for all relevant English articles published between 2001 and January 2018 and then performed forward citation tracking. The selected studies were RCTs and cohort studies with historical controls that explored the effect of prophylaxis with fluconazole in first 24 hours of life in VLBW infants on the incidence of IC colonization and mortality rates versus placebo or without fluconazole prophylaxis. Methodological quality was assessed using standardized methods, for instance, the Newcastle-Ottawa Scale was used to assess cohort studies [21]. The results were compared, and any questions or discrepancies were resolved through iteration and consensus.

The following data were extracted from the retrieved studies: baseline characteristics (authors, year of publication, number of patients, birth weight, and antifungal therapy characteristics) and outcomes of interest (incidence of proven IC, colonization, overall mortality, and IC-related mortality). A total of 174 articles were selected: 43 from PubMed, 32 from Scopus, 15 from Embase, 21 from Cochrane, and 63 from the manual search. Among the 174 articles, 84 were duplicated and 90 remained. Of them, 74 were excluded, including 29 reviews and expert commentary, 9 epidemiology of IC articles, 7 case reports, 5 case-control articles, 3 drug schedules and hygiene in NICU, 9 studies of pharmacokinetics of antifungal drugs, and 12 with unrelated titles from original articles describing antifungal prophylaxis use in VLBW infants. Sixteen studies were eligible for comparative evaluation (Fig. 1). As mentioned in Fig. 1, 3 studies that postponed starting prophylaxis more than 24 hours after birth were excluded. The primary outcome of this meta-analysis was mortality, while the secondary outcomes were colonization and IC.

For assessing the incidence of IC and mortality in VLBW and extremely low birth weight (ELBW) infants, fluconazole prophylaxis regimens used in 16 studies were reviewed [8-12,15, 22-31]. The study characteristics are mentioned in Table 1. All schedules consisted of 3- or 6-mg/kg fluconazole mostly for 4–6 weeks in 7 schedules for infants with a birth weight less than 1,500 g.

IC and/or mortality with or without colonization and IC-associated mortality were evaluated in 4,486 subjects in 9 RCTs and 8 cohort studies with 1,358 and 3,128 subjects, respectively, to compare the effect of prophylaxis with fluconazole and treatment with placebo or no prophylaxis in ELBW and VLBW infants. Table 2 shows the outcomes in addition the author’s name and year of the publication of every study. The meta-analysis included 4 groups of 2 or 3 studies each that had same study protocol, dose, frequency, and duration.

In every study, the risk ratios (RRs) for colonization, IC, IC-associated mortality, and mortality were determined by meta-analysis and the pooled RR was calculated to assess outcomes. The I² statistics and Cochrane Q were derived to estimate the fraction of variability between study RRs due to heterogeneity rather than chance. I² values of 25–75 and I² >75 were considered middle and high heterogeneity, respectively. In cases of heterogeneity, the random-effects model was used to determine the overall effect size [32]. Egger regression test and a funnel plot were used to assess publication bias. Statistical analyses were performed using comprehensive meta-analysis ver. 2.0 software (Biostat Inc., Englewood, CO, USA), and P values <0.05 were considered significant.

Mortality with or without IC was surveyed of the 7 RCTs and 9 cohort studies. Six studies showed significantly decreased mortality and/or IC-associated mortality in the prophylaxis groups. Rueda et al. [26] showed less IC-associated mortality in relation to all-cause death by prophylaxis year (P<0.05), while most deaths occurred in infants less than 1,000 g and neonates born at less than 31-week gestation. Kirpal et al. [31] reported that IC-associated mortality was significantly decreased with prophylaxis use in infants (Table 2). Regardless of schedule, the current meta-analysis showed that prophylaxis with fluconazole can significantly decrease the IC-associated mortality of VLBW infants weighing ≤1,000 g (P=0.011) (RR, 0.068; 95% confidence interval [CI], 0.009–0.543) (Figs. 2, 3) with different regimens and the administration of 3-mg fluconazole every 3 days in weeks 1 and 2, then every 2 days in weeks 3 and 4, and every day in weeks 5 and 6 decreased mortality rates significantly (P=0.023) (RR, 0.780; 95% CI, 0.629–0.966). The heterogeneity was not significant in these 2 outcomes as shown in Table 3 (group 2 for mortality: Q=3.20, degrees of freedom [df]=2, P=0.202, I²=37.486; group 2 for IC-associated mortality: Q=0.09, df=1, P=0.768, I²<0.001).

The colonization of Candida spp. in different locations (endotracheal secretion, nasopharynx, periumbilical region, perineum, gastric aspirate, skin) was evaluated in 7 studies (Table 2). All showed significantly decreased colonization, and the meta-analysis showed that 6 mg fluconazole administered every 3 days in week 1 and every day in the following 3 weeks can prevent colonization in VLBW infants (Fig. 4) (P<001; RR, 0.350; 95% CI, 0.223–0.551). Additionally, IC was decreased significantly on day 10, but 3 studies in the prophylaxis group versus placebo or the no prophylaxis group reported data in different groups, most of which were not suitable for analysis or the assessed schedules ineffectively prevented IC in the meta-analysis (Fig. 5). The heterogeneity was not significant in these outcomes as depicted in Table 3 (group 4 for colonization: Q=0.06, df=1, P=0.813, I²<0.001; group 4 for IC: Q=0.01, df=1, P=0.813, I²<0.001). The funnel plot of standard error by log RR is shown in Fig. 6. The results of Egger regression test were t=0.38, df=7.0, and P=0.71.